Hello Nur,

If the question of interaction is of importance to you, consider using multiple linear regression. The continuous IV goes into the model as is; the categorical IV will have to be recast into c - 1 dummy variates if it has more than two categories, C (if C =two, just code one level as 0 and the other level as 1). To evaluate for the presence of interaction, be sure to create c - 1 interaction terms to include in the model. The continuous DV would be entered as is.

You could compare model 1, with no interaction term(s) to model 2, which included the interaction term(s) along with the base IVs. If the R-squared difference was statistically significantly different from zero, then that would be evidence for the presence of interaction.

MLR does require some assumptions about the nature of the relationships (linear in form) and the nature of the errors (estimated by residuals) from the fitted model. Many statistical packages offer simple ways of evaluating the viability of these assumptions.

Good luck with your work.

Thank you David.

So to know if there is interaction, I just need to compare the R-squares of model 1 and model 2 only?

I am using RStudio by the way.

Maybe you could use an analysis of covariance. This is a linear model like this

dep = covariate + group (+ covariate*group)

group is the categorical independent variable

covariate is the continuous independent variable

In attachment an R script for such an analysis

As a matter of fact, one important assumption for an ANCOVA is that there is NO interaction between covariate*group. In those cases one can use the Johnson-Neyman procedure when heterogeneity of slopes. In attachment an R script for this cases.